A long-term localization and mapping system for autonomous inspection robots in large-scale environments using 3D...

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Inspection mobile robots with 3D LiDAR are crucial for substations but face limitations in large environments.
• Traditional LiDAR systems struggle with sensor degradation and dynamic objects, impacting map stability and robot localization accuracy.

Purpose of the Study:

• To develop a robust 3D LiDAR-based system for long-term localization and map maintenance in autonomous inspection robots.
• To overcome challenges of scale, sensor degradation, and dynamic environments for reliable robot operation.

Main Methods:

• A hierarchical system comprising a SLAM (Simultaneous Localization and Mapping) subsystem, a global localization subsystem, and a map maintenance subsystem.
• SLAM includes efficient local map representation, LiDAR odometry, global map optimization, and dense map generation using marching cubes.
• Global localization utilizes a novel ScanContext-ICP method for place recognition and pose initialization, followed by Iterative Closest Point (ICP) for refinement.
• Map maintenance tracks environmental changes, prioritizing stable features and filtering unstable points through temporal and segment-level analysis.

Main Results:

• The proposed system reduces average absolute position error by 30% and 38% compared to the state-of-the-art LOAM method on public datasets.
• Demonstrates superior accuracy and robustness, achieving a high SLAM ranking in real-world scenarios.
• The map maintenance system effectively improves localization accuracy by weighting stable features and filtering dynamic elements.

Conclusions:

• The developed 3D LiDAR-based system enables effective long-term localization and map maintenance for inspection mobile robots.
• The system's enhanced accuracy, robustness, and ability to handle dynamic environments make it suitable for autonomous deployment in critical infrastructure.
• The integration of hierarchical SLAM, advanced global localization, and intelligent map maintenance provides a significant advancement for robotic inspection applications.